FRI0679 PREDICTORS OF DEATH IN 3693 PATIENTS WITH RHEUMATOID ARTHRITIS FOLLOWED FOR UP TO 12 YEARS IN CLINICAL PRACTICE

Abstract

Background Patients with RA have shorter life expectancy than the background population. Severe disease, as reflected in high disease activity, comorbid conditions and functional disability have historically been associated with increased mortality (1-2). Mortality after the introduction of biological therapies has not been investigated in detail, and little is known about the predictive value of sociodemographic and lifestyle factors. Objectives To investigate the predictive value of a panel of potential predictors of death in a cohort of patients with RA followed for up to 12 years during the era of biological treatments. Methods Outpatients with RA were recruited consecutively between July 2006 and July 2007 and followed in routine care with prospective registrations in the DANBIO registry until death or August 30th 2018, whichever occurred first. Baseline variables considered to be potential predictors were: disease activity, disease duration, IgM-rheumatoid factor (IgM-RF), radiographic status (erosive disease yes/no) and medical therapy as well as patient-reported marital status, educational level, comorbid conditions, smoking, exercise, body mass index (BMI) and health assessment questionnaire (HAQ). Vital status and date of death were extracted from the Danish National Register. A cox proportional hazards model was used to estimate the hazard ratio for death for each of the potential predictors. Results 3693 patients were recruited at baseline, 75% women, 77% IgM-rheumatoid factor positive, 65% with erosive disease, median (IQR) age 62 years (52-71), disease duration 7 years (3-15), DAS28 3.0 (2.2-3.9), HAQ 0.63 (0.25-1.25). 20% received a biological disease modifying anti-rheumatic drug (DMARD), 71% received a synthetic DMARD and 9% received no DMARD. The median (IQR) duration of follow-up was 11 years (9-11); 1041 patients (28%) died during follow-up. 640 patients were excluded from the regression model due to missing baseline data; these individuals were slightly (median of 3 years) older than those who entered, but with similar disease duration and disease activity. All baseline variables were statistically significant predictors in univariable analyses. Table shows hazard ratio estimates in the multivariable model, which included 3053 patients; 762 (25%) deaths. IgM-RF positivity, higher HAQ score, glucocorticoid therapy, smoking (current and former) and two or more comorbid conditions were predictors of death. Low BMI was a borderline significant predictor for death. Female sex, weekly exercise and cohabiting decreased the risk of death. Conclusion In a large cohort of RA patients followed for a decade in the era of biological treatments, we identified strong clinical (high HAQ, comorbidity), treatment related (glucocorticoid last month), sociodemographic and lifestyle related (male sex, living alone, smoking, physical inactivity, low BMI) risk factors for death. In the effort to prevent a poor long term outcome in patients with RA, this study provides new insight into potentially modifiable baseline variables.